2 * Copyright (c) 1980, 1986, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)if.c 8.3 (Berkeley) 1/4/94
34 * $FreeBSD: src/sys/net/if.c,v 1.185 2004/03/13 02:35:03 brooks Exp $
35 * $DragonFly: src/sys/net/if.c,v 1.40 2005/06/15 19:29:30 joerg Exp $
38 #include "opt_compat.h"
39 #include "opt_inet6.h"
42 #include <sys/param.h>
43 #include <sys/malloc.h>
45 #include <sys/systm.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/socketops.h>
51 #include <sys/protosw.h>
52 #include <sys/kernel.h>
53 #include <sys/sockio.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
56 #include <sys/domain.h>
57 #include <sys/thread.h>
60 #include <net/if_arp.h>
61 #include <net/if_dl.h>
62 #include <net/if_types.h>
63 #include <net/if_var.h>
64 #include <net/ifq_var.h>
65 #include <net/radix.h>
66 #include <net/route.h>
67 #include <machine/stdarg.h>
69 #include <sys/thread2.h>
71 #if defined(INET) || defined(INET6)
73 #include <netinet/in.h>
74 #include <netinet/in_var.h>
75 #include <netinet/if_ether.h>
77 #include <machine/clock.h> /* XXX: temporal workaround for fxp issue */
78 #include <netinet6/in6_var.h>
79 #include <netinet6/in6_ifattach.h>
83 #if defined(COMPAT_43)
84 #include <emulation/43bsd/43bsd_socket.h>
85 #endif /* COMPAT_43 */
88 * Support for non-ALTQ interfaces.
90 static int ifq_classic_enqueue(struct ifaltq *, struct mbuf *,
91 struct altq_pktattr *);
93 ifq_classic_dequeue(struct ifaltq *, int);
94 static int ifq_classic_request(struct ifaltq *, int, void *);
97 * System initialization
100 static void if_attachdomain(void *);
101 static void if_attachdomain1(struct ifnet *);
102 static int ifconf (u_long, caddr_t, struct thread *);
103 static void ifinit (void *);
104 static void if_slowtimo (void *);
105 static void link_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
106 static int if_rtdel (struct radix_node *, void *);
108 SYSINIT(interfaces, SI_SUB_PROTO_IF, SI_ORDER_FIRST, ifinit, NULL)
110 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
111 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
112 MALLOC_DEFINE(M_CLONE, "clone", "interface cloning framework");
114 int ifqmaxlen = IFQ_MAXLEN;
115 struct ifnethead ifnet; /* depend on static init XXX */
119 * XXX: declare here to avoid to include many inet6 related files..
120 * should be more generalized?
122 extern void nd6_setmtu (struct ifnet *);
125 struct if_clone *if_clone_lookup (const char *, int *);
126 int if_clone_list (struct if_clonereq *);
128 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
129 int if_cloners_count;
131 struct callout if_slowtimo_timer;
134 * Network interface utility routines.
136 * Routines with ifa_ifwith* names take sockaddr *'s as
145 callout_init(&if_slowtimo_timer);
148 TAILQ_FOREACH(ifp, &ifnet, if_link) {
149 if (ifp->if_snd.ifq_maxlen == 0) {
150 if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
151 ifp->if_snd.ifq_maxlen = ifqmaxlen;
160 struct ifnet **ifindex2ifnet = NULL;
163 * Attach an interface to the
164 * list of "active" interfaces.
167 if_attach(struct ifnet *ifp)
169 unsigned socksize, ifasize;
170 int namelen, masklen;
171 struct sockaddr_dl *sdl;
175 static int if_indexlim = 8;
176 static boolean_t inited;
183 TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
184 ifp->if_index = ++if_index;
187 * The old code would work if the interface passed a pre-existing
188 * chain of ifaddrs to this code. We don't trust our callers to
189 * properly initialize the tailq, however, so we no longer allow
190 * this unlikely case.
192 TAILQ_INIT(&ifp->if_addrhead);
193 TAILQ_INIT(&ifp->if_prefixhead);
194 LIST_INIT(&ifp->if_multiaddrs);
195 getmicrotime(&ifp->if_lastchange);
196 if (ifindex2ifnet == NULL || if_index >= if_indexlim) {
202 /* grow ifindex2ifnet */
203 n = if_indexlim * sizeof(*q);
204 q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
206 bcopy(ifindex2ifnet, q, n/2);
207 free(ifindex2ifnet, M_IFADDR);
212 ifindex2ifnet[if_index] = ifp;
215 * create a Link Level name for this device
217 namelen = strlen(ifp->if_xname);
218 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
219 masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
220 socksize = masklen + ifp->if_addrlen;
221 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
222 if (socksize < sizeof(*sdl))
223 socksize = sizeof(*sdl);
224 socksize = ROUNDUP(socksize);
225 ifasize = sizeof(struct ifaddr) + 2 * socksize;
226 ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
227 sdl = (struct sockaddr_dl *)(ifa + 1);
228 sdl->sdl_len = socksize;
229 sdl->sdl_family = AF_LINK;
230 bcopy(ifp->if_xname, sdl->sdl_data, namelen);
231 sdl->sdl_nlen = namelen;
232 sdl->sdl_index = ifp->if_index;
233 sdl->sdl_type = ifp->if_type;
234 ifp->if_lladdr = ifa;
236 ifa->ifa_rtrequest = link_rtrequest;
237 ifa->ifa_addr = (struct sockaddr *)sdl;
238 sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
239 ifa->ifa_netmask = (struct sockaddr *)sdl;
240 sdl->sdl_len = masklen;
242 sdl->sdl_data[--namelen] = 0xff;
243 TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
245 EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
249 ifq->altq_disc = NULL;
250 ifq->altq_flags &= ALTQF_CANTCHANGE;
251 ifq->altq_tbr = NULL;
253 ifq_set_classic(ifq);
255 if (!SLIST_EMPTY(&domains))
256 if_attachdomain1(ifp);
258 /* Announce the interface. */
259 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
263 if_attachdomain(void *dummy)
268 TAILQ_FOREACH(ifp, &ifnet, if_list)
269 if_attachdomain1(ifp);
272 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST,
273 if_attachdomain, NULL);
276 if_attachdomain1(struct ifnet *ifp)
282 /* address family dependent data region */
283 bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
284 SLIST_FOREACH(dp, &domains, dom_next)
285 if (dp->dom_ifattach)
286 ifp->if_afdata[dp->dom_family] =
287 (*dp->dom_ifattach)(ifp);
292 * Detach an interface, removing it from the
293 * list of "active" interfaces.
296 if_detach(struct ifnet *ifp)
299 struct radix_node_head *rnh;
303 EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
306 * Remove routes and flush queues.
309 #ifdef DEVICE_POLLING
310 if (ifp->if_flags & IFF_POLLING)
311 ether_poll_deregister(ifp);
315 if (ifq_is_enabled(&ifp->if_snd))
316 altq_disable(&ifp->if_snd);
317 if (ifq_is_attached(&ifp->if_snd))
318 altq_detach(&ifp->if_snd);
321 * Clean up all addresses.
323 ifp->if_lladdr = NULL;
325 for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
326 ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
328 /* XXX: Ugly!! ad hoc just for INET */
329 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
330 struct ifaliasreq ifr;
332 bzero(&ifr, sizeof ifr);
333 ifr.ifra_addr = *ifa->ifa_addr;
334 if (ifa->ifa_dstaddr)
335 ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
336 if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
342 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
344 /* ifp_addrhead is already updated */
348 TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
354 * Remove all IPv4 kernel structures related to ifp.
361 * Remove all IPv6 kernel structs related to ifp. This should be done
362 * before removing routing entries below, since IPv6 interface direct
363 * routes are expected to be removed by the IPv6-specific kernel API.
364 * Otherwise, the kernel will detect some inconsistency and bark it.
370 * Delete all remaining routes using this interface
371 * Unfortuneatly the only way to do this is to slog through
372 * the entire routing table looking for routes which point
373 * to this interface...oh well...
375 for (i = 1; i <= AF_MAX; i++) {
376 if ((rnh = rt_tables[i]) == NULL)
378 rnh->rnh_walktree(rnh, if_rtdel, ifp);
381 /* Announce that the interface is gone. */
382 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
384 SLIST_FOREACH(dp, &domains, dom_next)
385 if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
386 (*dp->dom_ifdetach)(ifp,
387 ifp->if_afdata[dp->dom_family]);
390 * Remove interface from ifindex2ifp[] and maybe decrement if_index.
392 ifindex2ifnet[ifp->if_index] = NULL;
393 while (if_index > 0 && ifindex2ifnet[if_index] == NULL)
396 TAILQ_REMOVE(&ifnet, ifp, if_link);
401 * Delete Routes for a Network Interface
403 * Called for each routing entry via the rnh->rnh_walktree() call above
404 * to delete all route entries referencing a detaching network interface.
407 * rn pointer to node in the routing table
408 * arg argument passed to rnh->rnh_walktree() - detaching interface
412 * errno failed - reason indicated
416 if_rtdel(struct radix_node *rn, void *arg)
418 struct rtentry *rt = (struct rtentry *)rn;
419 struct ifnet *ifp = arg;
422 if (rt->rt_ifp == ifp) {
425 * Protect (sorta) against walktree recursion problems
428 if (!(rt->rt_flags & RTF_UP))
431 err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
432 rt_mask(rt), rt->rt_flags,
433 (struct rtentry **) NULL);
435 log(LOG_WARNING, "if_rtdel: error %d\n", err);
443 * Create a clone network interface.
446 if_clone_create(char *name, int len)
448 struct if_clone *ifc;
450 int wildcard, bytoff, bitoff;
454 ifc = if_clone_lookup(name, &unit);
458 if (ifunit(name) != NULL)
462 wildcard = (unit < 0);
464 * Find a free unit if none was given.
467 while (bytoff < ifc->ifc_bmlen &&
468 ifc->ifc_units[bytoff] == 0xff)
470 if (bytoff >= ifc->ifc_bmlen)
472 while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
474 unit = (bytoff << 3) + bitoff;
477 if (unit > ifc->ifc_maxunit)
480 err = (*ifc->ifc_create)(ifc, unit);
486 bitoff = unit - (bytoff << 3);
490 * Allocate the unit in the bitmap.
492 KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
493 ("%s: bit is already set", __func__));
494 ifc->ifc_units[bytoff] |= (1 << bitoff);
496 /* In the wildcard case, we need to update the name. */
498 for (dp = name; *dp != '\0'; dp++);
499 if (snprintf(dp, len - (dp-name), "%d", unit) >
500 len - (dp-name) - 1) {
502 * This can only be a programmer error and
503 * there's no straightforward way to recover if
506 panic("if_clone_create(): interface name too long");
511 EVENTHANDLER_INVOKE(if_clone_event, ifc);
517 * Destroy a clone network interface.
520 if_clone_destroy(const char *name)
522 struct if_clone *ifc;
527 ifc = if_clone_lookup(name, &unit);
531 if (unit < ifc->ifc_minifs)
538 if (ifc->ifc_destroy == NULL)
541 (*ifc->ifc_destroy)(ifp);
544 * Compute offset in the bitmap and deallocate the unit.
547 bitoff = unit - (bytoff << 3);
548 KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
549 ("%s: bit is already cleared", __func__));
550 ifc->ifc_units[bytoff] &= ~(1 << bitoff);
555 * Look up a network interface cloner.
558 if_clone_lookup(const char *name, int *unitp)
560 struct if_clone *ifc;
564 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
565 for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
566 if (ifc->ifc_name[i] != *cp)
571 ifc = LIST_NEXT(ifc, ifc_list);
575 return ((struct if_clone *)NULL);
581 for (i = 0; *cp != '\0'; cp++) {
582 if (*cp < '0' || *cp > '9') {
583 /* Bogus unit number. */
586 i = (i * 10) + (*cp - '0');
596 * Register a network interface cloner.
599 if_clone_attach(struct if_clone *ifc)
606 KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
607 ("%s: %s requested more units then allowed (%d > %d)",
608 __func__, ifc->ifc_name, ifc->ifc_minifs,
609 ifc->ifc_maxunit + 1));
611 * Compute bitmap size and allocate it.
613 maxclone = ifc->ifc_maxunit + 1;
615 if ((len << 3) < maxclone)
617 ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
618 ifc->ifc_bmlen = len;
620 LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
623 for (unit = 0; unit < ifc->ifc_minifs; unit++) {
624 err = (*ifc->ifc_create)(ifc, unit);
626 ("%s: failed to create required interface %s%d",
627 __func__, ifc->ifc_name, unit));
629 /* Allocate the unit in the bitmap. */
631 bitoff = unit - (bytoff << 3);
632 ifc->ifc_units[bytoff] |= (1 << bitoff);
637 * Unregister a network interface cloner.
640 if_clone_detach(struct if_clone *ifc)
643 LIST_REMOVE(ifc, ifc_list);
644 free(ifc->ifc_units, M_CLONE);
649 * Provide list of interface cloners to userspace.
652 if_clone_list(struct if_clonereq *ifcr)
654 char outbuf[IFNAMSIZ], *dst;
655 struct if_clone *ifc;
656 int count, error = 0;
658 ifcr->ifcr_total = if_cloners_count;
659 if ((dst = ifcr->ifcr_buffer) == NULL) {
660 /* Just asking how many there are. */
664 if (ifcr->ifcr_count < 0)
667 count = (if_cloners_count < ifcr->ifcr_count) ?
668 if_cloners_count : ifcr->ifcr_count;
670 for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
671 ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
672 strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
673 error = copyout(outbuf, dst, IFNAMSIZ);
682 * Locate an interface based on a complete address.
685 ifa_ifwithaddr(struct sockaddr *addr)
690 TAILQ_FOREACH(ifp, &ifnet, if_link)
691 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
692 if (ifa->ifa_addr->sa_family != addr->sa_family)
694 if (sa_equal(addr, ifa->ifa_addr))
696 if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
697 /* IPv6 doesn't have broadcast */
698 ifa->ifa_broadaddr->sa_len != 0 &&
699 sa_equal(ifa->ifa_broadaddr, addr))
702 return ((struct ifaddr *)NULL);
705 * Locate the point to point interface with a given destination address.
708 ifa_ifwithdstaddr(struct sockaddr *addr)
713 TAILQ_FOREACH(ifp, &ifnet, if_link)
714 if (ifp->if_flags & IFF_POINTOPOINT)
715 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
716 if (ifa->ifa_addr->sa_family != addr->sa_family)
718 if (ifa->ifa_dstaddr &&
719 sa_equal(addr, ifa->ifa_dstaddr))
722 return ((struct ifaddr *)NULL);
726 * Find an interface on a specific network. If many, choice
727 * is most specific found.
730 ifa_ifwithnet(struct sockaddr *addr)
734 struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
735 u_int af = addr->sa_family;
736 char *addr_data = addr->sa_data, *cplim;
739 * AF_LINK addresses can be looked up directly by their index number,
740 * so do that if we can.
743 struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
745 if (sdl->sdl_index && sdl->sdl_index <= if_index)
746 return (ifindex2ifnet[sdl->sdl_index]->if_lladdr);
750 * Scan though each interface, looking for ones that have
751 * addresses in this address family.
753 TAILQ_FOREACH(ifp, &ifnet, if_link) {
754 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
755 char *cp, *cp2, *cp3;
757 if (ifa->ifa_addr->sa_family != af)
759 if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
761 * This is a bit broken as it doesn't
762 * take into account that the remote end may
763 * be a single node in the network we are
765 * The trouble is that we don't know the
766 * netmask for the remote end.
768 if (ifa->ifa_dstaddr != NULL &&
769 sa_equal(addr, ifa->ifa_dstaddr))
773 * if we have a special address handler,
774 * then use it instead of the generic one.
776 if (ifa->ifa_claim_addr) {
777 if ((*ifa->ifa_claim_addr)(ifa, addr)) {
785 * Scan all the bits in the ifa's address.
786 * If a bit dissagrees with what we are
787 * looking for, mask it with the netmask
788 * to see if it really matters.
791 if (ifa->ifa_netmask == 0)
794 cp2 = ifa->ifa_addr->sa_data;
795 cp3 = ifa->ifa_netmask->sa_data;
796 cplim = ifa->ifa_netmask->sa_len +
797 (char *)ifa->ifa_netmask;
799 if ((*cp++ ^ *cp2++) & *cp3++)
800 goto next; /* next address! */
802 * If the netmask of what we just found
803 * is more specific than what we had before
804 * (if we had one) then remember the new one
805 * before continuing to search
806 * for an even better one.
808 if (ifa_maybe == 0 ||
809 rn_refines((char *)ifa->ifa_netmask,
810 (char *)ifa_maybe->ifa_netmask))
819 * Find an interface address specific to an interface best matching
823 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
826 char *cp, *cp2, *cp3;
828 struct ifaddr *ifa_maybe = 0;
829 u_int af = addr->sa_family;
833 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
834 if (ifa->ifa_addr->sa_family != af)
838 if (ifa->ifa_netmask == NULL) {
839 if (sa_equal(addr, ifa->ifa_addr) ||
840 (ifa->ifa_dstaddr != NULL &&
841 sa_equal(addr, ifa->ifa_dstaddr)))
845 if (ifp->if_flags & IFF_POINTOPOINT) {
846 if (sa_equal(addr, ifa->ifa_dstaddr))
850 cp2 = ifa->ifa_addr->sa_data;
851 cp3 = ifa->ifa_netmask->sa_data;
852 cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
853 for (; cp3 < cplim; cp3++)
854 if ((*cp++ ^ *cp2++) & *cp3)
863 #include <net/route.h>
866 * Default action when installing a route with a Link Level gateway.
867 * Lookup an appropriate real ifa to point to.
868 * This should be moved to /sys/net/link.c eventually.
871 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
874 struct sockaddr *dst;
877 if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
878 (ifp = ifa->ifa_ifp) == NULL || (dst = rt_key(rt)) == NULL)
880 ifa = ifaof_ifpforaddr(dst, ifp);
885 if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
886 ifa->ifa_rtrequest(cmd, rt, info);
891 * Mark an interface down and notify protocols of
893 * NOTE: must be called at splnet or eqivalent.
896 if_unroute(struct ifnet *ifp, int flag, int fam)
900 ifp->if_flags &= ~flag;
901 getmicrotime(&ifp->if_lastchange);
902 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
903 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
904 pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
905 ifq_purge(&ifp->if_snd);
910 * Mark an interface up and notify protocols of
912 * NOTE: must be called at splnet or eqivalent.
915 if_route(struct ifnet *ifp, int flag, int fam)
919 ifp->if_flags |= flag;
920 getmicrotime(&ifp->if_lastchange);
921 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
922 if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
923 pfctlinput(PRC_IFUP, ifa->ifa_addr);
931 * Mark an interface down and notify protocols of the transition. An
932 * interface going down is also considered to be a synchronizing event.
933 * We must ensure that all packet processing related to the interface
934 * has completed before we return so e.g. the caller can free the ifnet
935 * structure that the mbufs may be referencing.
937 * NOTE: must be called at splnet or eqivalent.
940 if_down(struct ifnet *ifp)
942 if_unroute(ifp, IFF_UP, AF_UNSPEC);
943 netmsg_service_sync();
947 * Mark an interface up and notify protocols of
949 * NOTE: must be called at splnet or eqivalent.
952 if_up(struct ifnet *ifp)
955 if_route(ifp, IFF_UP, AF_UNSPEC);
959 * Handle interface watchdog timer routines. Called
960 * from softclock, we decrement timers (if set) and
961 * call the appropriate interface routine on expiration.
964 if_slowtimo(void *arg)
970 TAILQ_FOREACH(ifp, &ifnet, if_link) {
971 if (ifp->if_timer == 0 || --ifp->if_timer)
973 if (ifp->if_watchdog)
974 (*ifp->if_watchdog)(ifp);
979 callout_reset(&if_slowtimo_timer, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
983 * Map interface name to
984 * interface structure pointer.
987 ifunit(const char *name)
992 * Search all the interfaces for this name/number
995 TAILQ_FOREACH(ifp, &ifnet, if_link) {
996 if (strncmp(ifp->if_xname, name, IFNAMSIZ) == 0)
1004 * Map interface name in a sockaddr_dl to
1005 * interface structure pointer.
1008 if_withname(struct sockaddr *sa)
1010 char ifname[IFNAMSIZ+1];
1011 struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
1013 if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
1014 (sdl->sdl_nlen > IFNAMSIZ) )
1018 * ifunit wants a null-terminated name. It may not be null-terminated
1019 * in the sockaddr. We don't want to change the caller's sockaddr,
1020 * and there might not be room to put the trailing null anyway, so we
1021 * make a local copy that we know we can null terminate safely.
1024 bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
1025 ifname[sdl->sdl_nlen] = '\0';
1026 return ifunit(ifname);
1034 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1042 size_t namelen, onamelen;
1043 char new_name[IFNAMSIZ];
1045 struct sockaddr_dl *sdl;
1051 return (ifconf(cmd, data, td));
1053 ifr = (struct ifreq *)data;
1058 if ((error = suser(td)) != 0)
1060 return ((cmd == SIOCIFCREATE) ?
1061 if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1062 if_clone_destroy(ifr->ifr_name));
1064 case SIOCIFGCLONERS:
1065 return (if_clone_list((struct if_clonereq *)data));
1068 ifp = ifunit(ifr->ifr_name);
1074 ifr->ifr_flags = ifp->if_flags;
1075 ifr->ifr_flagshigh = ifp->if_flags >> 16;
1079 ifr->ifr_reqcap = ifp->if_capabilities;
1080 ifr->ifr_curcap = ifp->if_capenable;
1084 ifr->ifr_metric = ifp->if_metric;
1088 ifr->ifr_mtu = ifp->if_mtu;
1092 ifr->ifr_phys = ifp->if_physical;
1099 new_flags = (ifr->ifr_flags & 0xffff) |
1100 (ifr->ifr_flagshigh << 16);
1101 if (ifp->if_flags & IFF_SMART) {
1102 /* Smart drivers twiddle their own routes */
1103 } else if (ifp->if_flags & IFF_UP &&
1104 (new_flags & IFF_UP) == 0) {
1108 } else if (new_flags & IFF_UP &&
1109 (ifp->if_flags & IFF_UP) == 0) {
1115 #ifdef DEVICE_POLLING
1116 if ((new_flags ^ ifp->if_flags) & IFF_POLLING) {
1117 if (new_flags & IFF_POLLING) {
1118 ether_poll_register(ifp);
1120 ether_poll_deregister(ifp);
1125 ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1126 (new_flags &~ IFF_CANTCHANGE);
1127 if (new_flags & IFF_PPROMISC) {
1128 /* Permanently promiscuous mode requested */
1129 ifp->if_flags |= IFF_PROMISC;
1130 } else if (ifp->if_pcount == 0) {
1131 ifp->if_flags &= ~IFF_PROMISC;
1134 (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1135 getmicrotime(&ifp->if_lastchange);
1142 if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1144 (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1151 error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1154 if (new_name[0] == '\0')
1156 if (ifunit(new_name) != NULL)
1159 EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
1161 /* Announce the departure of the interface. */
1162 rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1164 strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1165 ifa = TAILQ_FIRST(&ifp->if_addrhead);
1166 /* XXX IFA_LOCK(ifa); */
1167 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1168 namelen = strlen(new_name);
1169 onamelen = sdl->sdl_nlen;
1171 * Move the address if needed. This is safe because we
1172 * allocate space for a name of length IFNAMSIZ when we
1173 * create this in if_attach().
1175 if (namelen != onamelen) {
1176 bcopy(sdl->sdl_data + onamelen,
1177 sdl->sdl_data + namelen, sdl->sdl_alen);
1179 bcopy(new_name, sdl->sdl_data, namelen);
1180 sdl->sdl_nlen = namelen;
1181 sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1182 bzero(sdl->sdl_data, onamelen);
1183 while (namelen != 0)
1184 sdl->sdl_data[--namelen] = 0xff;
1185 /* XXX IFA_UNLOCK(ifa) */
1187 EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
1189 /* Announce the return of the interface. */
1190 rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1197 ifp->if_metric = ifr->ifr_metric;
1198 getmicrotime(&ifp->if_lastchange);
1207 error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1209 getmicrotime(&ifp->if_lastchange);
1214 u_long oldmtu = ifp->if_mtu;
1219 if (ifp->if_ioctl == NULL)
1220 return (EOPNOTSUPP);
1221 if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1223 error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1225 getmicrotime(&ifp->if_lastchange);
1229 * If the link MTU changed, do network layer specific procedure.
1231 if (ifp->if_mtu != oldmtu) {
1245 /* Don't allow group membership on non-multicast interfaces. */
1246 if ((ifp->if_flags & IFF_MULTICAST) == 0)
1249 /* Don't let users screw up protocols' entries. */
1250 if (ifr->ifr_addr.sa_family != AF_LINK)
1253 if (cmd == SIOCADDMULTI) {
1254 struct ifmultiaddr *ifma;
1255 error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1257 error = if_delmulti(ifp, &ifr->ifr_addr);
1260 getmicrotime(&ifp->if_lastchange);
1263 case SIOCSIFPHYADDR:
1264 case SIOCDIFPHYADDR:
1266 case SIOCSIFPHYADDR_IN6:
1268 case SIOCSLIFPHYADDR:
1270 case SIOCSIFGENERIC:
1274 if (ifp->if_ioctl == 0)
1275 return (EOPNOTSUPP);
1276 error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1278 getmicrotime(&ifp->if_lastchange);
1282 ifs = (struct ifstat *)data;
1283 ifs->ascii[0] = '\0';
1285 case SIOCGIFPSRCADDR:
1286 case SIOCGIFPDSTADDR:
1287 case SIOCGLIFPHYADDR:
1289 case SIOCGIFGENERIC:
1290 if (ifp->if_ioctl == 0)
1291 return (EOPNOTSUPP);
1292 return ((*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred));
1298 return if_setlladdr(ifp,
1299 ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1302 oif_flags = ifp->if_flags;
1303 if (so->so_proto == 0)
1304 return (EOPNOTSUPP);
1306 error = so_pru_control(so, cmd, data, ifp, td);
1313 case SIOCSIFDSTADDR:
1315 case SIOCSIFBRDADDR:
1316 case SIOCSIFNETMASK:
1317 #if BYTE_ORDER != BIG_ENDIAN
1318 if (ifr->ifr_addr.sa_family == 0 &&
1319 ifr->ifr_addr.sa_len < 16) {
1320 ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1321 ifr->ifr_addr.sa_len = 16;
1324 if (ifr->ifr_addr.sa_len == 0)
1325 ifr->ifr_addr.sa_len = 16;
1333 case OSIOCGIFDSTADDR:
1334 cmd = SIOCGIFDSTADDR;
1337 case OSIOCGIFBRDADDR:
1338 cmd = SIOCGIFBRDADDR;
1341 case OSIOCGIFNETMASK:
1342 cmd = SIOCGIFNETMASK;
1344 error = so_pru_control(so, cmd, data, ifp, td);
1348 case OSIOCGIFDSTADDR:
1349 case OSIOCGIFBRDADDR:
1350 case OSIOCGIFNETMASK:
1351 *(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1355 #endif /* COMPAT_43 */
1357 if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1359 DELAY(100);/* XXX: temporary workaround for fxp issue*/
1360 if (ifp->if_flags & IFF_UP) {
1374 * Set/clear promiscuous mode on interface ifp based on the truth value
1375 * of pswitch. The calls are reference counted so that only the first
1376 * "on" request actually has an effect, as does the final "off" request.
1377 * Results are undefined if the "off" and "on" requests are not matched.
1380 ifpromisc(struct ifnet *ifp, int pswitch)
1386 oldflags = ifp->if_flags;
1387 if (ifp->if_flags & IFF_PPROMISC) {
1388 /* Do nothing if device is in permanently promiscuous mode */
1389 ifp->if_pcount += pswitch ? 1 : -1;
1394 * If the device is not configured up, we cannot put it in
1397 if ((ifp->if_flags & IFF_UP) == 0)
1399 if (ifp->if_pcount++ != 0)
1401 ifp->if_flags |= IFF_PROMISC;
1402 log(LOG_INFO, "%s: promiscuous mode enabled\n",
1405 if (--ifp->if_pcount > 0)
1407 ifp->if_flags &= ~IFF_PROMISC;
1408 log(LOG_INFO, "%s: promiscuous mode disabled\n",
1411 ifr.ifr_flags = ifp->if_flags;
1412 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1413 error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1414 (struct ucred *)NULL);
1418 ifp->if_flags = oldflags;
1423 * Return interface configuration
1424 * of system. List may be used
1425 * in later ioctl's (above) to get
1426 * other information.
1429 ifconf(u_long cmd, caddr_t data, struct thread *td)
1431 struct ifconf *ifc = (struct ifconf *)data;
1434 struct sockaddr *sa;
1435 struct ifreq ifr, *ifrp;
1436 int space = ifc->ifc_len, error = 0;
1438 ifrp = ifc->ifc_req;
1439 TAILQ_FOREACH(ifp, &ifnet, if_link) {
1442 if (space <= sizeof ifr)
1446 * Zero the stack declared structure first to prevent
1447 * memory disclosure.
1449 bzero(&ifr, sizeof(ifr));
1450 if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1451 >= sizeof(ifr.ifr_name)) {
1452 error = ENAMETOOLONG;
1457 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1458 if (space <= sizeof ifr)
1461 if (td->td_proc->p_ucred->cr_prison &&
1466 if (cmd == OSIOCGIFCONF) {
1467 struct osockaddr *osa =
1468 (struct osockaddr *)&ifr.ifr_addr;
1470 osa->sa_family = sa->sa_family;
1471 error = copyout(&ifr, ifrp, sizeof ifr);
1475 if (sa->sa_len <= sizeof(*sa)) {
1477 error = copyout(&ifr, ifrp, sizeof ifr);
1480 if (space < (sizeof ifr) + sa->sa_len -
1483 space -= sa->sa_len - sizeof(*sa);
1484 error = copyout(&ifr, ifrp,
1485 sizeof ifr.ifr_name);
1487 error = copyout(sa, &ifrp->ifr_addr,
1489 ifrp = (struct ifreq *)
1490 (sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1494 space -= sizeof ifr;
1499 bzero(&ifr.ifr_addr, sizeof ifr.ifr_addr);
1500 error = copyout(&ifr, ifrp, sizeof ifr);
1503 space -= sizeof ifr;
1507 ifc->ifc_len -= space;
1512 * Just like if_promisc(), but for all-multicast-reception mode.
1515 if_allmulti(struct ifnet *ifp, int onswitch)
1523 if (ifp->if_amcount++ == 0) {
1524 ifp->if_flags |= IFF_ALLMULTI;
1525 ifr.ifr_flags = ifp->if_flags;
1526 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1527 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1528 (struct ucred *)NULL);
1531 if (ifp->if_amcount > 1) {
1534 ifp->if_amcount = 0;
1535 ifp->if_flags &= ~IFF_ALLMULTI;
1536 ifr.ifr_flags = ifp->if_flags;
1537 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1538 error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1539 (struct ucred *)NULL);
1551 * Add a multicast listenership to the interface in question.
1552 * The link layer provides a routine which converts
1556 struct ifnet *ifp, /* interface to manipulate */
1557 struct sockaddr *sa, /* address to add */
1558 struct ifmultiaddr **retifma)
1560 struct sockaddr *llsa, *dupsa;
1562 struct ifmultiaddr *ifma;
1565 * If the matching multicast address already exists
1566 * then don't add a new one, just add a reference
1568 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1569 if (sa_equal(sa, ifma->ifma_addr)) {
1570 ifma->ifma_refcount++;
1578 * Give the link layer a chance to accept/reject it, and also
1579 * find out which AF_LINK address this maps to, if it isn't one
1582 if (ifp->if_resolvemulti) {
1583 error = ifp->if_resolvemulti(ifp, &llsa, sa);
1584 if (error) return error;
1589 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1590 MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1591 bcopy(sa, dupsa, sa->sa_len);
1593 ifma->ifma_addr = dupsa;
1594 ifma->ifma_lladdr = llsa;
1595 ifma->ifma_ifp = ifp;
1596 ifma->ifma_refcount = 1;
1597 ifma->ifma_protospec = 0;
1598 rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1601 * Some network interfaces can scan the address list at
1602 * interrupt time; lock them out.
1605 LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1610 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1611 if (sa_equal(ifma->ifma_addr, llsa))
1615 ifma->ifma_refcount++;
1617 MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1618 M_IFMADDR, M_WAITOK);
1619 MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1620 M_IFMADDR, M_WAITOK);
1621 bcopy(llsa, dupsa, llsa->sa_len);
1622 ifma->ifma_addr = dupsa;
1623 ifma->ifma_ifp = ifp;
1624 ifma->ifma_refcount = 1;
1626 LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1631 * We are certain we have added something, so call down to the
1632 * interface to let them know about it.
1635 ifp->if_ioctl(ifp, SIOCADDMULTI, 0, (struct ucred *)NULL);
1642 * Remove a reference to a multicast address on this interface. Yell
1643 * if the request does not match an existing membership.
1646 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
1648 struct ifmultiaddr *ifma;
1650 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1651 if (sa_equal(sa, ifma->ifma_addr))
1656 if (ifma->ifma_refcount > 1) {
1657 ifma->ifma_refcount--;
1661 rt_newmaddrmsg(RTM_DELMADDR, ifma);
1662 sa = ifma->ifma_lladdr;
1664 LIST_REMOVE(ifma, ifma_link);
1666 * Make sure the interface driver is notified
1667 * in the case of a link layer mcast group being left.
1669 if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1670 ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1672 free(ifma->ifma_addr, M_IFMADDR);
1673 free(ifma, M_IFMADDR);
1678 * Now look for the link-layer address which corresponds to
1679 * this network address. It had been squirreled away in
1680 * ifma->ifma_lladdr for this purpose (so we don't have
1681 * to call ifp->if_resolvemulti() again), and we saved that
1682 * value in sa above. If some nasty deleted the
1683 * link-layer address out from underneath us, we can deal because
1684 * the address we stored was is not the same as the one which was
1685 * in the record for the link-layer address. (So we don't complain
1688 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1689 if (sa_equal(sa, ifma->ifma_addr))
1694 if (ifma->ifma_refcount > 1) {
1695 ifma->ifma_refcount--;
1700 LIST_REMOVE(ifma, ifma_link);
1701 ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1703 free(ifma->ifma_addr, M_IFMADDR);
1704 free(sa, M_IFMADDR);
1705 free(ifma, M_IFMADDR);
1711 * Set the link layer address on an interface.
1713 * At this time we only support certain types of interfaces,
1714 * and we don't allow the length of the address to change.
1717 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1719 struct sockaddr_dl *sdl;
1723 sdl = IF_LLSOCKADDR(ifp);
1726 if (len != sdl->sdl_alen) /* don't allow length to change */
1728 switch (ifp->if_type) {
1729 case IFT_ETHER: /* these types use struct arpcom */
1734 bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1737 bcopy(lladdr, LLADDR(sdl), len);
1743 * If the interface is already up, we need
1744 * to re-init it in order to reprogram its
1747 if ((ifp->if_flags & IFF_UP) != 0) {
1748 ifp->if_flags &= ~IFF_UP;
1749 ifr.ifr_flags = ifp->if_flags;
1750 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1751 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1752 (struct ucred *)NULL);
1753 ifp->if_flags |= IFF_UP;
1754 ifr.ifr_flags = ifp->if_flags;
1755 ifr.ifr_flagshigh = ifp->if_flags >> 16;
1756 (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1757 (struct ucred *)NULL);
1760 * Also send gratuitous ARPs to notify other nodes about
1761 * the address change.
1763 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1764 if (ifa->ifa_addr != NULL &&
1765 ifa->ifa_addr->sa_family == AF_INET)
1766 arp_ifinit(ifp, ifa);
1773 struct ifmultiaddr *
1774 ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp)
1776 struct ifmultiaddr *ifma;
1778 LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1779 if (sa_equal(ifma->ifma_addr, sa))
1786 * The name argument must be a pointer to storage which will last as
1787 * long as the interface does. For physical devices, the result of
1788 * device_get_name(dev) is a good choice and for pseudo-devices a
1789 * static string works well.
1792 if_initname(struct ifnet *ifp, const char *name, int unit)
1794 ifp->if_dname = name;
1795 ifp->if_dunit = unit;
1796 if (unit != IF_DUNIT_NONE)
1797 snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1799 strlcpy(ifp->if_xname, name, IFNAMSIZ);
1803 if_printf(struct ifnet *ifp, const char *fmt, ...)
1808 retval = printf("%s: ", ifp->if_xname);
1809 __va_start(ap, fmt);
1810 retval += vprintf(fmt, ap);
1815 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1816 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1819 ifq_set_classic(struct ifaltq *ifq)
1821 ifq->altq_enqueue = ifq_classic_enqueue;
1822 ifq->altq_dequeue = ifq_classic_dequeue;
1823 ifq->altq_request = ifq_classic_request;
1827 ifq_classic_enqueue(struct ifaltq *ifq, struct mbuf *m,
1828 struct altq_pktattr *pa __unused)
1831 if (IF_QFULL(ifq)) {
1842 static struct mbuf *
1843 ifq_classic_dequeue(struct ifaltq *ifq, int op)
1856 panic("unsupported ALTQ dequeue op: %d", op);
1863 ifq_classic_request(struct ifaltq *ifq, int req, void *arg)
1871 panic("unspported ALTQ request: %d", req);